INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIAL AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

When imaging in scattering tissue, nonresonant two-photon excitation (TPE) fluorescence microscopy reportedly is superior to its single-photon excitation (SPE) counterpart (Denk et al., 1990, 1996). The quadratic dependence of the excitation on the illumination intensity confines the excitation to the focal region so that the mere generation of fluorescence in a raster fashion establishes a three-dimensional image. Importantly, ballistic and scattered emitted photons contribute equally well to the total signal. This allows the collection of a large fraction of fluorescence photons by external detectors resulting in a much higher collection efficiency. Moreover, TPE of visible dyes requires excitation wavelengths in the near-infrared where scattering coefficients are lower (Svaasand and Ellingsen, 1983) and, with rare exceptions, linear absorption is negligible. Weak endogenous absorption and spatial confinement of TPE inflicts considerably lower phototoxic stress than SPE, which is invaluable for live-cell imaging (Wokosin et al., 1996; Maiti et al., 1997). However, the nonlinear nature of TPE requires much higher intensities. The high intensities may well induce photodamage and irreversible sample modifications.

To provide an estimate for the intensities and power levels involved, let us first review briefly the power and intensity levels in SPE and TPE microscopy, assuming typical wavelengths of _exc = 450 nm and _exc = 870 nm, respectively, and a numerical aperture NA = 1.0. The SPE rate is given by k₁ = ₁I/(hc/) whereby h is Planck's constant, and c the speed of light (Lakowicz, 1983). For a typical cross section of ₁ = 10¹⁶ cm², an intensity I = 200 kW/cm² is sufficient to reach k₁ = 5 × 10⁷ s¹, which is close to saturation by ground state depletion. This intensity corresponds well to a continuous wave average power of ~140 µW at the focus. To obtain the same rate, k₂ = ₂[I/(hc/)]², with nonresonant TPE, for a typical TPE cross section of ₂ = 10⁴⁹ cm⁴s (Xu and Webb, 1996), a much higher intensity of 5.4 GW/cm² is required. Given the same lenses and the wavelength needed for TPE is _exc = 870 nm, this intensity corresponds to an average power of 12 W at the focus.

Although efficient continuous wave TPE microscopy is feasible for highly concentrated staining and large TPE cross sections (Booth and Hell, 1998; Hell et al., 1998), TPE microscopy usually requires pulsed lasers such as the mode-locked Ti:Sapphire, providing pulses of   100 fs or   1 ps pulse duration, at a pulse repetition rate f  80 MHz. For pulsed illumination, the n-photon absorption rate is given by

(1)

where I(t) denotes the intensity. Because most power detectors are too slow to follow I(t), the excitation rate is often expressed in terms of the time-averaged power value measured by the detector. For 1/f, the evaluation of the integral in the paraxial focusing approximation gives

(2)

where and denote the time-averaged intensity and time-averaged power, respectively. The constant g⁽ⁿ⁾ is the time-zero n-order temporal coherence. For hyperbolic-secant-squared shaped pulses, g⁽²⁾ = 0.59. When considering photodamage issues, it is important to bear in mind that the time-averaged values are not determining the absorption rate alone. They are normally used because they can be easily measured by a (slow) detector. The relationship between the peak intensity I_peak and the time-average intensity is I_peak = (f)¹. The proportionality of k_n to (I_peak)ⁿ¹ reflects the fact that (only) nonlinear (n > 1) processes benefit from high peak intensities.

Compared to the CW TPE case, pulsing reduces the required time-averaged power by a factor of (f)^1/2 but, at the same time, increases the peak intensity by the same factor. By rearranging Eq. 2, one can calculate that, for the focusing conditions and assumed TPE rate k₂ = 5 × 10⁷ s¹, pulsed-mode TPE imaging with pulses of  = 1 ps, and 100 fs would keep the time-averaged power down to 140 mW and 45 mW, respectively, at the expense of involving peak intensities of I_peak = 800 GW/cm² in the 1 ps, and even higher intensities of 2.5 TW/cm², in the 100 fs mode.

Standard Ti:Sapphire systems emit between 0.5 and 2 W of time-averaged power. Thus, the laser usually provides ample power to saturate nearly all dyes by TPE. In fact, the available power would allow an increase in the focal intensity by another order of magnitude. Evidently, dyes with 100 times lower cross sections could still be excited with saturating intensities. However, Eq. 2 indicates that the high peak intensities of >1 TW/cm² foster nonlinear (n > 1) processes, some of which might be detrimental to the fluorophore or the sample. Such intensities can overwhelm the light specimen interaction before saturation by ground-state-depletion is reached.

Such evidence for irreversible fluorophore damage induced by TPE has indeed been found in single-molecule experiments (Eggeling et al., 1998). Photodamage in TPE microscopy has also been reported by several researchers in the field, often as a sudden localized onset of increased luminescence (König et al., 1999; Straub and Hell, 1998). It is reasonable to expect the limiting mechanisms to vary with the sample and fluorophore. In fact, such limits may well be a combination of linear and nonlinear processes, with each of them prevailing at a given power and intensity range. It will be important to establish the dependence of the damage on the excitation intensity as a function of the power order n of the (non-)linearity.

According to Eq. 2, photodamage by linear absorption (n = 1) depends solely on the average intensity and not on the peak intensity I_peak. If linear photodamage dominated, safe TPE would be possible only for high transient intensities such as can be obtained by minimizing the pulse width. If cubic or higher order (n  3) processes would dominate photodamage, increasing I_peak by reducing the pulse width would be disadvantageous. In the special case of a quadratic dependence, the photodamage and TPE rate will just be proportional to each other, and the attainable TPE rate would not depend on the pulse width.

As in the earlier stages of TPE microscopy, linear absorption was considered to be the limiting factor, so, keeping the pulses as short as possible was considered of particular importance (Denk et al., 1995). Prechirping units were applied to compensate for group velocity dispersion in the microscope (Soeller and Canell, 1996; Brakenhoff et al., 1995), so that the pulse could reach the specimen at its shortest possible duration (~100 fs). However, recent two-photon imaging with increased through picosecond (7 ps) pulsed (Bewersdorf and Hell, 1998) and CW-excitation ( = 200 mW) (Hell et al., 1998) revealed nonlinear optical effects as the major limitations and indicated that the reduction of the pulse duration is of secondary importance.

In a more recent study, a large population of live unstained Chinese hamster ovarian cells were exposed to Ti:Sapphire laser illumination and the impairment of cell replication quantified as a measure for photodamage (König et al., 1999). For pulses ranging from 2.2 ps down to 240 fs, the photodamage followed explicitly a quadratic power law; this rendered the pulse width irrelevant for optimizing for least photophysical damage. Whereas, for the 240-fs pulses, an average power of  = 7.3 mW was sufficient to impair cell viability, for the 2.2-ps pulses, an average power of  = 20 mW was required to produce the same effect. Taking into account the NA = 1.35, the corresponding focal intensities I_peak were 0.35 TW/cm² for the femtosecond and 0.1 TW/cm² for the picosecond pulses. Although these results are very valuable, the use of chirped laser pulses and the complexity of the damage mechanisms does not allow a general conclusion for other applications.

Due to its high penetration depth, TPE has been particularly successful in the imaging of Ca²⁺ dynamics in small neuronal compartments of live brain tissue (Svoboda et al., 1996; Yuste and Denk, 1995; Schiller et al., 1998; Koester and Sakmann, 1998; Denk et al., 1995; Yuste et al., 1999). Although the relevance of characterizing phototoxic effects has been recognized, information about the relationship between excitation power and intensity and photodamage in TPE microscopy is scarce. Finding out the optimal operational parameters for multiphoton imaging is of paramount importance to the proper use of this powerful imaging mode. Therefore, in our study, we investigated the limiting factors of TPE-Ca²⁺ imaging in rat brain slices using various fluorescent Ca²⁺ indicators and pulse widths. In particular, we provide a quantitative description of the observed photodamage and investigate the order of nonlinearity of the damage.

	MATERIAL AND METHODS

TOP ABSTRACT INTRODUCTION MATERIAL AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Slice preparation

Acute neocortical slices were prepared from 13 to 14-day-old Wistar rats as described previously (Markram et al., 1997). Layer 5 pyramidal neurons in the somatosensory cortical area were identified using gradient-contrast (Dodt et al., 1998) infrared videomicroscopy. The bath solution contained 125 mM NaCl, 25 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 1 mM MgCl₂, 25 mM Glucose, and 2 mM CaCl₂ (Biometra, Goettingen, Germany). The bath temperature was 32-34°C. All experiments were performed on basal dendrites 30-80 µm away from the soma. To keep the attenuation of excitation laser power comparable, only structures 30-70 µm below the slice surface were selected.

Electrophysiology

Cells were filled with dyes by whole-cell recording. Patch pipettes were filled with 115 µM K-gluconate, 20 µM KCl, 10 µM HEPES, 4 µM ATP-Mg, 10 µM phosphocreatine, 0.3 µM GTP, and a dye, which was either Calcium Green-1 (CG-1, 100 µM), Oregon Green 488 BAPTA-1 (OGB-1, 100, 200, or 500 µM), Oregon Green 488 BAPTA-2 (OGB-2, 500 µM), Magnesium Green (MG, 500 µM), or Fluorescein (500 µM). All dyes were purchased from Molecular Probes (Eugene, OR). Dyes were allowed to equilibrate for 20 min after establishing whole-cell recording before experiments started. Electrical recordings were made with a patch-clamp amplifier (EPC-7, List-Electronics, Darmstadt, Germany) operated in the current clamp mode. Initial access resistances were <6 M. Action potentials (AP) were evoked by brief current injection (1 nA for 4-7 ms) into the soma.

Two-photon laser scanning microscopy

A modified galvanometer scanning unit (TCS 4D, Leica Microsystems, Heidelberg, Germany) was adapted to an upright microscope (BX50WI, Olympus Optical Co., Tokyo, Japan) equipped with a 60× objective (LUMPlanFL 60xW0.9IR, Olympus). For excitation femtosecond laser pulses and picosecond pulses that were nearly transform-limited in the focal plane were used. In a few cases, we also used femtosecond pulses chirped to picoseconds. The femtosecond laser pulses originated from a Ti:Sa-Laser (MIRA 900F, Coherent, Santa Clara, CA) pumped by a large-frame argon ion laser (Sabre ML14, Coherent). The nearly transform-limited picosecond pulses originated from a picosecond Ti:Sa-Laser (Tsunami, Spectra-Physics, Mountain View, CA). Both lasers were operated at 870 ± 5 nm. Insertion of a mirror allowed a convenient change between the lasers during the experiment. When imaging in the line scan mode, signals were collected during both scan directions. External detectors were placed behind the objective and the condensor for signal collection. Transmission- and epifluorescence signals were recorded by photomultiplier tubes (PMT) (R6357, Hamamatsu Photonics, Herrsching, Germany) and digitized into 8-bit values. The signals were averaged off-line. In control experiments, a bandpass filter (525 ± 25 nm, Leica Microsystems) was inserted into the epifluorescence pathway.

Analysis of calcium fluorescence transients

Fluorescence line scan images were analyzed using dedicated software. A line was scanned every 2.27 ms. Pixels of two lines between two positions enclosing the dendrite examined were averaged to obtain one time point. This resulted in a temporal resolution of 4.54 ms. Stimulation protocols began 150 ms after the start of the line scan (512 lines). Before stimulation, fluorescence was averaged for 100 ms to obtain the basal fluorescence, F₀. A region distant from any indicator-containing structure was chosen for determination of the background fluorescence, F_B, which was subtracted. Fluorescence traces F(t) of a Ca²⁺ transient evoked by an AP were fitted with a single exponential using a least-square fit routine (IGOR, Wave- Metrics, Lake Oswego, OR). This fit rendered the amplitude of the fluorescence increase A_F and decay time constant . Relative fluorescence changes were calculated as F(t)/F₀ = (F(t)  F₀)/(F₀  F_B). A_R was the amplitude of the relative fluorescence transient.

Pulse compression and diagnostics

Pulse width control

View larger version (12K): [in this window] [in a new window]   FIGURE 1   Determination of the pulse width by second-order autocorrelation. (A) F Background free autocorrelation of pulses of a femtosecond mode Ti:Sapphire laser before entering the microscope; FWHM was 112 ± 6 fs. Assuming Gaussian pulses, the pulse width amounts to 79 fs. (B) Interferometric autocorrelation in the objective focal plane showing an FWHM of 114 ± 5 fs (pulse width 81 fs). (C) Background free autocorrelation of chirped pulses. To broaden femtosecond pulses to the picosecond regime, 36 cm of SF6 glass was inserted into the optical path. The background free autocorrelation signal had a width of 1.96 ps (pulse width 1.4 ps). (D) Autocorrelation of the pulses in the focal plane shows a strong chirp induced by the SF6-glass (FWHM  2 ps).

Interferometric autocorrelation

Background-free autocorrelation

Laser power determination

Determination of the photodamage rate r at different pulse lengths

Illumination protocol

d

Determination of r

Comparing picosecond and femtosecond laser pulses in single cells

(3)

(4)

Detection scheme

We used detectors placed at three different positions. One detector was the confocal, descanned detector with the pinhole removed. A second detector for capturing nondescanned epifluorescence was placed behind the water immersion objective (LumPlanFl 63W0.9IR, NA 0.9, Olympus). A third, external detector was placed behind the high NA = 1.4 oil condensor (see Fig. 2). We used the same type of PMT (R6357, Hamamatsu), supplied with the same voltage for all three detector positions. Because the sensitive area of the side-window PMTs is only 8 × 24 mm, they required additional optics to achieve the scan-fields desired. For demagnification, we placed lenses with focal length of f = 12 mm (for the transmission detector) and f = 20 mm (epifluorescence external detector) in the corresponding planes. The virtual scanfield of the condensor PMT depends on the aperture (Fig. 2). In the system used, the virtual scanfield was about 1.2 mm. External detection through the NA = 1.4 oil condensor and the NA = 0.9 water immersion objective and descanned detection were compared using a fluorescence solution layer (Fluorescein). The number of photons and gain was calculated from the histograms to exclude variation of gain and dark current between the PMTs. The signal collected by the descanned detector was about 70% of the signal collected by the epifluorescence nondescanned detector. With the same fluorescence solution, the signal collected by the transmission detector was 215 ± 65% (mean ± SD, n = 18) of the descanned epifluorescence signal. Because the situation might be different in highly scattering tissue, we also tested dye-filled (OGB-1, 200 µM) neurons in brain slices. Here, the number of photons collected by the transmission external detection compared to the descanned epifluorescence detection ranged up to 450% (n = 2).

View larger version (22K): [in this window] [in a new window]   FIGURE 2   Nondescanned transmission collection. The optical elements including the condenser lenses are shown. The short-pass filter allowed IR-video imaging of the brain slices. The field stop used corresponded to a virtual scan-field of 1.2 mm.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIAL AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Calcium dynamics of single back-propagating AP and cumulative photodamage

One major use of two-photon laser scanning microscopy is calcium imaging in small compartments of cells in highly scattering tissue. Therefore, we investigated the fluorescence dynamics in layer 5 neocortical neurons in rat brain slices filled with a fluorescent calcium indicator dye. Upon evoking an AP, the physiologically relevant transient increase of intradendritic free calcium was observed as a transient increase in fluorescence of the calcium indicator (Fig. 3 A). A single exponential function satisfactorily fitted the decay. The decay times varied between 50 and 400 ms, depending on the buffering by the dye (Helmchen et al., 1996). The calcium fluorescence dynamics can be described by the following set of parameters: the basal fluorescence F₀, the amplitude of the fluorescence change A_F, and the decay time constant . Changes in these parameters during prolonged laser light exposure were elected as measures of photodamage.

View larger version (17K): [in this window] [in a new window]   FIGURE 3   Changes in calcium dynamics at low excitation rates. (A) Somatic current injection (1 nA for 6 ms) evoked a back-propagating AP that elicited a calcium transient in the dendrites of a layer 5 pyramidal cell. The calcium transient was reported by a calcium indicator, OGB-1 (200 µM). The trace presented is an average of five sweeps. The parameters describing this calcium fluorescence transient are basal fluorescence F0, amplitudes AR (and AF for the relative fluorescence change), and decay time constant  (see Methods). (B-D) At an average laser power of 6.4 mW, changes in the parameters during 50 s of light exposure were observed. In the example shown, the basal fluorescence increased by nearly 300%, the relative amplitude AR decreased. The absolute amplitude AF changed only little. (E) The decay time constant increased from ~210 to ~320 ms. Different cell as in A-D, average laser power was 7.2 mW.

The relative fluorescence change F/F(t) was calculated and the amplitude A_R was monitored during up to 400 s of laser light exposure. Whereas the basal fluorescence F₀ and decay time  increased with time, the amplitude of the relative fluorescence change A_R decreased (see Fig. 3, B-E). At low average laser powers, that is <3-7 mW (depending on the depth of the cell within the slice) for femtosecond laser pulses and <8-24 mW for picosecond laser pulses, all changes were observed to be linear with exposure time. The changes were restricted to the illuminated area (see Fig. 4 A); they were cumulative and irreversible. This type of changes is further referred to as cumulative photodamage.

View larger version (58K): [in this window] [in a new window]   FIGURE 4   Photodamage at various excitation rates. (A) Image of a basal dendrite of a layer 5 pyramidal cell filled with OGB-1 (100 µM) showing a stationary and highly localized increase in basal fluorescence caused by repetitive line scans using femtosecond laser pulses (75 ± 4 fs). The broken line indicates the position of the line scans performed before recording a stack of 10 images. Line scans were recorded with an average laser power of  = 5.3 mW, the xy-image with a lower power (about 3 mW). (B) Image of a basal dendrite showing vesicular, dye-filled structures (arrows) caused by repetitive line scan imaging with high average laser power ( = 9.5-11.8 mW).

To rule out loading effects, the average laser power and pulse length were randomly varied. Also, the results did not correlate with loading time or distance to the soma. Most experiments were performed on dendrites close to the soma, where the dye concentration reached equilibrium within a few minutes (Helmchen et al., 1996). This is well below the time we waited after establishing the whole-cell recording before starting the experiments (>20 min).

The increase of the basal fluorescence was also observable without back-propagating APs (OGB-2, n = 2 and MG, n = 18 with n representing the number of trials) and when a dye was used that was not sensitive to calcium (Fluorescein, n = 8). The decrease of A_R could be explained by the increased basal fluorescence, because the amplitude of the absolute fluorescence change of the calcium fluorescence transient, A_F, remained constant (Fig. 3 D). The change of the basal fluorescence was observable with all dyes used. Qualitatively, these changes did also occur when intermitted laser light exposure instead of continuous exposure was used (e.g., 1 s exposure every 10-20 s). These characteristics allowed us to use the rates of change as quantitative measures of photodamage. The changes in basal fluorescence exhibited the highest S/N-ratio; therefore, the rate of relative change in basal fluorescence, r (see Methods), was taken as a measure for cumulative photodamage.

Photodamage at high excitation rates

Aside from the cumulative photodamage described above, a broad spectrum of phototoxic effects was observed. When calcium-sensitive dyes were used, sharp increases in F₀ occasionally occurred in the absence of any electrophysiological signal. These sudden increases were irregular, depended nonlinearly on the applied exposure time, and were not localized to illuminated parts of the cell. In fact, they extended some micrometers along the dendrite. These effects were regularly observed at  > 8 and  > 30 mW for femtosecond and picosecond pulses, respectively. Increased photobleaching and other morphological alterations of spines and dendrites occurred at these laser powers, such as the formation of vesicular dye-filled structures (Fig. 4 B). For determination of r, we restricted the laser power intensity to lower values so that only the linear changes in basal fluorescence were observed. Experiments in which one of the other types of artifacts were observed were excluded from data. The damage rate r ranged up to 1.6%/s for  < 8 mW (femtosecond laser pulses) and  < 30 mW (picosecond laser pulses).

Emission spectrum

A possible explanation for an increased level of fluorescence is the generation of a new chromophore from cellular compounds. Therefore, the emission spectrum was examined for possible changes. In some experiments, we compared the fluorescence from throughout the detectable spectrum (350 nm < _em < 600 nm) with the fluorescence emitted at 525 ± 25 nm and calculated the corresponding photodamage rates r. The dyes used have their emission maximum in the wavelength interval around 510-540 nm. The bandpass filter was used only for the epifluorescence signal. In this way, it was possible to compare fluorescence at 525 nm with the detectable fluorescence signal. The comparison was carried out for OGB-1 (100, 200, and 500 µM) and MG (500 µM). No significant difference between the two rates r could be observed. On average, the r at 525 nm was 95 ± 6% (n = 15, ± SE) of the r calculated from the total fluorescence. This indicates that the increase in basal fluorescence is not caused by the generation of a new chromophore. Furthermore, no changes could be detected in the background fluorescence F_B (all dyes and laser powers, n > 100) supporting this point of view.

Dependence of the photodamage rate r on dye concentration

Cells at similar depth at close quarters were patched and filled by whole-cell recording with the same indicator at different concentrations (OGB-1, 200 and 500 µM). In this way, it was possible to choose cell structures at very close quarters (<15 µm) and at equal depth containing different concentrations of dye. At a given average laser power, and thus, a constant TPE rate, the damage rate r was determined in both structures, and thus, at different dye concentrations. On average, we found r(500 µM)/r(200 µM) = 104 ± 15% (n = 7, ± SE). Despite the 2.5-fold difference in concentration, no significant difference between the damage rates was found. This indicates that the measure of cumulative photodamage used did not depend on dye concentration; it suggests that r is a robust measure for photodamage.

Photodamage and diffusion inhibition

The slower kinetics of the calcium fluorescence transients suggest changes in mobility in damaged areas. Therefore, we tested the diffusion properties of the dye upon evoking photodamage. For this purpose, cells (n = 4) were filled with fluorescein (500 µM) by whole-cell recording. At a location on a basal dendrite (e.g., the location marked by * in Fig. 5), a significant change in basal fluorescence was induced by illumination. The patch pipette with fluorescein was withdrawn and the cell repatched, this time without any dye in the patch pipette. Washout of the dye molecules lead to a decrease of fluorescence in the cell. At a location nearby (<10 µm distance) and proximal to the damaged site, the fluorescence decreased exponentially with a time constant close to that anticipated for the dye to diffuse back through the dendrite into the pipette ( = 30-70 s, depending on the distance to the soma). However, in comparison to that, the fluorescence from within the damaged area decreased very slowly with a decay time >100-fold larger ( > 6000 s, in all four cells examined). This drastic decrease in mobility indicates that a part of the dye molecules was bound and was no longer able to diffuse. Indeed, even the very slow decrease may have been caused mainly by photobleaching that is normally masked by diffusion of unbleached dye molecules into the observed area.

View larger version (34K): [in this window] [in a new window]   FIGURE 5   Changes in diffusion after eliciting photodamage. (A) Image of a layer 5 pyramidal cell filled with fluorescein (500 µM) by whole-cell recording. At a certain place on a basal dendrite (indicated by *) a line was scanned for 65 s at an average laser power (10 mW, 80-fs pulses) that induced an increase in basal fluorescence (image recorded before this procedure). (B) Decrease in fluorescence by washout after repatching. The cell was repatched with a pipette containing no dye. The decrease in fluorescence due to the washout of the dye was monitored. In the damaged area (*) the decay of fluorescence was very slow, indicating that diffusion was inhibited. The decay time  = 6900 s was more than 100-fold larger than in the unaffected area (1). (C) Proximal (1) to the damaged area, the washout of the dye occurred with a regular diffusion constant ( = 65). (D) The decay of fluorescence was monitored also at a place (2) distal to the damaged area, where the dye had to diffuse through the damaged area (*) for washout. The decay time constant ( = 288 ms) was 4 times higher than in (1).

An important detail was revealed by another observation. In two experiments (of four cells), the decay of fluorescence distal to the damaged site was significantly slower than its proximal counterpart: 288 versus 65 and 150 versus 72 ms, in spite of close (10 µm) proximity between the locations. That means that unaffected dye diffused more slowly through damaged areas. This difference is consistent with the hypothesis that damaged areas were changed structurally in correspondence with the observed alterations in calcium fluorescence kinetics.

Dependence of the photodamage rate r on pulse length

Changing the pulse length is well suited for estimating the predominant order of the damage process at a given TPE rate. By adjusting the pulse length and peak intensity so that · I_peak² is kept constant, the TPE rate can be kept constant while higher- and lower-order processes are being changed. Conversely, if the damage rate does not depend on · I_peak², then the damaging process must be a two-photon mechanism. The influence of the pulse duration on the damage rate r was tested (see Methods) by altering the pulse width from 75 fs to 0.8-3.2 ps. The excitation rates used were kept low to avoid the additional, irregular photodamaging mechanisms occurring at higher excitation rates. At these excitation rates, r ranged up to 0.25%/s. For 22 cells, r was measured as a function of the average laser power for femtosecond and picosecond laser pulses within a small power interval. For equal r, the exponent x could be derived by means of Eq. 4. No difference was found between chirped (n = 5) and nearly transform-limited laser pulses (n = 17) regarding the calculated exponents; therefore all data were pooled. We found an average exponent of x = 1.997 ± 0.38 (n = 22, ± SD, Fig. 6), indicating a two-photon-induced photodamaging mechanism at low excitation rates.

View larger version (8K): [in this window] [in a new window]   FIGURE 6   Dependence of cumulative photodamage on pulse length. Using Eq. 4, the order of excitation for the damage process was calculated by comparing similar damage rates r in a pulse length interval from 75 fs to 3.2 ps at low TPE rates. The histogram shows the results of 22 measurements with a mean of 1.997 ± 0.38 (mean ± SD).

Dependence of the photodamage rate r on excitation rate

It has been speculated that higher-order processes like three-photon excitation contribute to photodamage at the peak intensity and power levels used for imaging. One might ask if, at the low excitation rates used, potential third-order components were masked by noise. Therefore, we tested the dependence of the photodamage on the excitation rate for femto- and picosecond laser pulses, separately. For a given pulse length, r was measured at different average laser powers for femtosecond (Fig. 7 A) or picosecond laser pulses (Fig. 7 B). The dependence on was fitted to determine the exponent x (see Methods) using Eq. 3. We tested 25 cells. Since no differences between the calculated exponents using different calcium dyes (MG: n = 10; OGB-1: 100 µM, n = 2; 200 µM, n = 3; 500 µM, n = 2; OGB-2: 500 µM, n = 3) and chirped (n = 11) and nearly transform-limited (n = 13) picosecond laser pulses were observed, all data were pooled. For femtosecond laser pulses, the average exponent was x = 2.5 ± 1.0 (mean ± SD, n = 16). For picosecond laser pulses, the average exponent was x = 2.2 ± 1.0 (mean ± SD, n = 12). Since calcium indicators were used in these experiments, it was possible that the changes in basal fluorescence resulted from changes in resting calcium levels. Therefore, control experiments with fluorescein (500 µM) were performed. The average exponent was 2.5 ± 0.5 (mean ± SD, n = 5) for femtosecond laser pulses. This indicates that the results did not depend on whether a calcium indicator or an unspecific dye was used.

View larger version (16K): [in this window] [in a new window]   FIGURE 7   Dependence of cumulative photodamage on average laser power. (A) The graph shows four measurements of the rate of relative change r (in %) of basal fluorescence at different time-averaged powers  = 4.8-11.8 mW (dots). Measurements were carried out in the same cell and at different locations in similar depths on a basal dendrite of a layer 5 pyramidal cell filled with MG (500 µM). Resting membrane potential was 67 mV and pulse length was 78 ± 5 fs. The continuous line shows a fit using Eq. 3, yielding the function r = 0.0004%/s · (/mW)3.35. The upper left insert shows a histogram of evaluated exponents for femtosecond laser pulses. (B) The graph shows measurements of r (in %) at different time-averaged powers  = 8.5-15.8 mW (dots) from a different cell. The dye used was MG (500 µM), resting membrane potential was 64 mV and pulse length was ~1.4 ps. The continuous line shows a fit using Eq. 3, yielding the function r = 0.00042%/s · (/mW)2.18. The upper left insert shows the histogram of exponents for picosecond laser pulses.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIAL AND METHODS RESULTS DISCUSSION APPENDIX REFERENCES

To fully exploit the benefits of two-photon laser scanning microscopy, detailed knowledge is required about power limitations and potential photodamage. Various criteria have been proposed to quantify photodamage, strongly depending on the particular biological application of multiphoton imaging. In multiphoton imaging of cultured cells, photodamage has been related to the decrease in cell viability and quantified through the observation of the metabolic state (Sako et al., 1997; König et al., 1999). In the case of calcium imaging of neuronal cells, photodamage can be observed as nonphysiological functional and morphological changes. At low excitation rates, changes in AP-evoked calcium fluorescence transients occurred. The observed changes are independent of the occurrence of APs, clearly localized to areas exposed to the laser, proportional to the time elapsed, and irreversible. The fact that the basal fluorescence increased linearly with time allowed us to define the rate of basal fluorescence increase as a dependable measure of photodamage. The damage rates are presented in terms of average laser power. The quantification in terms of TPE rates would be useful but compromized by several assumptions, such as scattering in brain slices, refractive mismatch-induced abberations, etc. Nevertheless, we want to give an estimate of the orders of magnitudes involved; this is done in the Appendix.

A possible explanation of the increase in the basal level of fluorescence is that some dye molecules undergo changes binding them to the cellular matrix or membranes. Hence the basal fluorescence increased, because the mobile fraction of dye equilibrated again, raising the absolute concentration of dye. In contrast, the absolute fluorescence increase A_F evoked by an AP remained constant, indicating that the bound dye molecularly altered and no longer reported to calcium changes. This is supported by a previous report for another dye, fura-2, where a similar intermediate photoproduct with altered characteristics was found (Becker and Fay, 1987). Based on these assumptions, one can calculate that the maximal number of photons a dye molecule emits before it binds is <10⁶ (see Appendix). In our case, the generation of a new chromophore is unlikely because we did not observe a significant spectral difference between the fluorescence from the photodamaged and undamaged areas. We also did not observe an unexpected onset of fluorescence in cell structures that did not contain dye molecules. The steady increase in basal fluorescence used for quantification of cumulative photodamage should not be confused with the abrupt increases of basal fluorescence observed at higher intensities when a calcium-sensitive dye was used. The latter might be caused by calcium entering through a leaking external or internal membrane.

In addition, we found a hampered diffusion of dye molecules through damaged areas and a slower decay of the calcium fluorescence transients. This might be explained by a reduced mobility of the dye molecules. If this interpretation is right, it indicates much more severe changes than only the binding of dye molecules to intracellular structures. One possible interpretation is that the intracellular environment was heavily changed and became viscous, presumably by the binding of many proteins to each other and the cytoskeleton, comparable to what happens in fixation procedures. Such a severe change would impair the physiology of the affected cell structure because proteins cannot fulfil their functions anymore.

Several observations indicate the physiological relevance of the measure used for the cumulative photodamage. The increase in the decay time of the calcium fluorescence transients and the reduced diffusion through and out of damaged cell parts suggest that the photodamage affects physiological parameters that are independent of the fluorophore. Furthermore, the laser intensities of excitation correlate to those reported where photodamage in unstained cells has been examined (König et al., 1999).

The investigation of the role of the pulse length is of particular interest because it reveals the dependence of the photodamage mechanism on the excitation intensity. If we just consider the TPE rate (Eq. 2), we find that, to achieve the same two-photon fluorescence as the femtosecond (₁ = 100 fs) illumination, the picosecond (₂ = 2 ps) pulse train requires a time-averaged power , or time-averaged intensity that is higher by a factor of   4.5 but a peak power, or peak intensity, that is lower by the same factor. (Note that, in Eq. 2, the TPE rate is proportional to the product of average and peak power). Our experiments show that this variation in pulse length, average, and peak power can really be carried out and that calcium signals can be equally well observed with picosecond laser pulses (Fig. 8). This is in accordance with previous findings (Hänninen et al., 1994; Hell et al., 1998; Bewersdorf and Hell, 1998; Schönle and Hell, 1998; Jenei et al., 1999; König et al., 1999) and sufficient to prove that the mechanism responsible for the cumulative photodamage is not caused by a linear absorption mechanism such as heating through single-photon absorption. In fact, at low excitation rates, the measurements revealed a quadratic exponent, x = 2.0 ± 0.4, indicating that cumulative photodamage is dominated largely by two-photon absorption. Hence, for a given photodamage rate r, the femtosecond pulses involved a higher peak power I_peak in connection with a lower average power, whereas the picosecond pulses used a higher average power in conjunction with a lower peak power I_peak, just as suggested by Eq. 2. As a good rule of thumb, the rate of photodamage induced is determined by the product of the squared intensities times the pulse length: · I_peak²  (P · NA²)²/. This relationship indicates that higher time-averaged power can be afforded if the aperture is lower or the pulse length is longer. In other words, for low excitation rates, the photodamage is just proportional to the two-photon fluorescence rate.

View larger version (60K): [in this window] [in a new window]   FIGURE 8   Comparison of femto- and picosecond mode imaging. Projections of stacks of 25 images of a layer 5 pyramidal neuron filled with OGB-2 (500 µM) by somatic whole-cell recording; excitation wavelength was  = 870 nm. (A) Pulse length used was  = 1.4 ps and average laser power was  30 mW. (B) Pulse length used was  = 74 + 4 fs (nearly transform-limited) and average lasers power was   7 mW. Similar two-photon laser scanning rates were achieved in both cases.

Although, at low excitation rates, we found a quadratic nature for the damaging mechanism, at higher rates, we measured exponents >2.0. This might indicate that higher-order effects that are too weak to be observed at low intensities may become more important at higher excitation rates. The data presented are indeed consistent with this assumption, but this interpretation is difficult to test due to the onset of additional mechanisms of damage. These irregular artifacts observed at higher excitation rates could not be quantified and might have affected the estimation of the exponents when evaluating the dependence of the photodamage rate r on the average laser power.

In summary, the aim of this work was to clarify to what extent possible light-specimen interactions influence physiologically significant parameters in multiphoton calcium imaging. Furthermore, we investigated the role of peak intensity, excitation rate, and pulse length in photodamage. For low excitation rates, we established the rate of increase of basal fluorescence as an index of photodamage. This parameter enabled us to examine the order of the damaging process by changing the pulse length. The observed exponent of 2.0 ± 0.4 indicates two-photon absorption as the predominant damaging mechanism. This means that photodamage in TPE imaging is proportional to the signal and, thus, that signal can only be obtained at the expense of photodamage. When imaging over a longer time, low excitation rates must be used to avoid too heavy changes. This results unfortunately in a lower S/N-ratio. One way to overcome this is to optimize the collection efficiency (see Methods: Detection scheme). One should keep this in mind when applying multiphoton imaging to live cells. For low excitation rates, the results indicate that, within a window of 75 fs to 3.2 ps, the pulse width is irrelevant for two-photon calcium imaging.

Nevertheless, the use of multiphoton imaging provides several advantages over confocal microscopy in highly scattering tissue. The possibility of using ballistic photons for signal generation allows the capture of a much larger fraction of the fluorescence photons than in confocal microscopy, and the spatial confinement of excitation inflicts considerably less phototoxic stress to the cell. By carefully adjusting safe intensity levels for imaging, these benefits can be fully exploited.